Thank you for the wonderful pipeline. Apologies in advance, I have tried to find this answer in your code/paper, but still dont understand.

When I get a a summaryDT from summary(fit, doLRT = condition) -- I was looking at the correlation between z-scores and -log(Pr(>Chisq)) for all the various components (C, D, H, etc).

I am confused as to why the continuous z-score and my Hurdle-pvalue are highly correlated (>0.97 across a bunch of cell types and contrasts), but the logFC z-scores are ~0.2 with the hurdle p-values. If I understand from the paper you are summing the Wald/LRT test statistics to derive a single p-value (H-component) with and without condition. Conceptually I guess I don't understand why the hurdle p-value should be so closely associated with only the continuous component, but not the logFC z-score?

Thanks again.